The Role of Ethnic Origin, Phenotype, and Physical Attractiveness in Hiring Processes: Evidence from a Conjoint Experiment with Spanish Recruiters

The GEMM Discrimination Study (Polavieja et al. 2023) documented sizable phenotypic discrimination in Germany and the Netherlands but weaker effects in Spain. A key design choice in GEMM—portraying all applicants as physically attractive—may have attenuated phenotype penalties if attractiveness mitigates bias, as suggested by U.S. evidence (Monk et al. 2021). We directly probe this possibility in Spain, a setting where photographs are routinely included in résumés, enabling a realistic test of how phenotype and attractiveness, alongside ethnicity and gender, shape early-stage hiring. We field a preregistered, paired-profile forced-choice conjoint experiment among Spanish HR professionals (N = 351; recruiters with current hiring authority n = 173, our primary analytic sample). Ten profile attributes are randomized (ethnic origin signaled by names; phenotype and attractiveness via standardized facial images; gender; education; university; productivity signal; English; communication skills; employment status), with citizenship, experience (3 years), and age (27) held constant. Occupation (pharmacy technician—high client contact vs. software engineer—low client contact) is randomized between subjects. The design yields 7,020 profile evaluations and supports estimation of AMCEs, probability-scale AMEs, and AMIEs. In the full HR sample, AMCEs are generally small; Black phenotype shows a modest negative effect (−2.22 pp; p = .07). Focusing on recruiters and translating effects into selection probabilities (AMEs) under empirically relevant conditions, we find: Spanish-origin (vs. Moroccan-origin) names increase interview selection by 3.54 pp; above-average attractiveness raises selection by 3.59 pp; the Black phenotype estimate is negative but statistically indistinguishable from zero; and, unexpectedly, unemployed (vs. employed) candidates are 8.71 pp more likely to be selected, consistent with recruiters valuing immediate availability at first screen. Formal tests detect no statistically significant two- or three-way interactions (phenotype×ethnicity, phenotype×attractiveness, or phenotype×attractiveness×gender), nor evidence that attractiveness matters more in client-facing roles. Substantively, these results underscore that culturally coded signals (names) and appearance (attractiveness) move shortlisting probabilities in Spain, while phenotypic penalties—if present—are weak in this recruiter context. Methodologically, the study provides, to our knowledge, the first recruiter-focused Spanish conjoint that integrates facial imagery, and the first experimental test outside the U.S. of Monk et al.’s interaction claim—finding no support and thereby sharpening its scope conditions. Because recruiters are gatekeepers, even 3–4 pp shifts at the shortlist stage can compound across large applicant pools. Our preregistered, image-enhanced conjoint thus advances discrimination research by combining design-based identification (AMCEs) with decision-relevant probability shifts (AMEs) to illuminate where, and for whom, early-stage hiring inequalities arise.